diff --git a/mokuro/manga_page_ocr.py b/mokuro/manga_page_ocr.py
index c64d012..abb9ae2 100644
--- a/mokuro/manga_page_ocr.py
+++ b/mokuro/manga_page_ocr.py
@@ -9,6 +9,7 @@
 from mokuro import __version__
 from mokuro.cache import cache
 from mokuro.utils import imread
+import torch
 
 
 class InvalidImage(Exception):
@@ -35,9 +36,11 @@ def __init__(
         self.disable_ocr = disable_ocr
 
         if not self.disable_ocr:
-            logger.info("Initializing text detector")
+            cuda = torch.cuda.is_available()
+            device = 'cuda' if cuda and not force_cpu else 'cpu'
+            logger.info(f"Initializing text detector, using device {device}")
             self.text_detector = TextDetector(
-                model_path=cache.comic_text_detector, input_size=detector_input_size, device="cpu", act="leaky"
+                model_path=cache.comic_text_detector, input_size=detector_input_size, device=device, act="leaky"
             )
             self.mocr = MangaOcr(pretrained_model_name_or_path, force_cpu)